1 /* 2 * Copyright (c) 2012 The WebRTC project authors. All Rights Reserved. 3 * 4 * Use of this source code is governed by a BSD-style license 5 * that can be found in the LICENSE file in the root of the source 6 * tree. An additional intellectual property rights grant can be found 7 * in the file PATENTS. All contributing project authors may 8 * be found in the AUTHORS file in the root of the source tree. 9 */ 10 11 #ifndef MODULES_AUDIO_CODING_NETEQ_MERGE_H_ 12 #define MODULES_AUDIO_CODING_NETEQ_MERGE_H_ 13 14 #include "modules/audio_coding/neteq/audio_multi_vector.h" 15 #include "rtc_base/constructor_magic.h" 16 17 namespace webrtc { 18 19 // Forward declarations. 20 class Expand; 21 class SyncBuffer; 22 23 // This class handles the transition from expansion to normal operation. 24 // When a packet is not available for decoding when needed, the expand operation 25 // is called to generate extrapolation data. If the missing packet arrives, 26 // i.e., it was just delayed, it can be decoded and appended directly to the 27 // end of the expanded data (thanks to how the Expand class operates). However, 28 // if a later packet arrives instead, the loss is a fact, and the new data must 29 // be stitched together with the end of the expanded data. This stitching is 30 // what the Merge class does. 31 class Merge { 32 public: 33 Merge(int fs_hz, 34 size_t num_channels, 35 Expand* expand, 36 SyncBuffer* sync_buffer); 37 virtual ~Merge(); 38 39 // The main method to produce the audio data. The decoded data is supplied in 40 // |input|, having |input_length| samples in total for all channels 41 // (interleaved). The result is written to |output|. The number of channels 42 // allocated in |output| defines the number of channels that will be used when 43 // de-interleaving |input|. 44 virtual size_t Process(int16_t* input, 45 size_t input_length, 46 AudioMultiVector* output); 47 48 virtual size_t RequiredFutureSamples(); 49 50 protected: 51 const int fs_hz_; 52 const size_t num_channels_; 53 54 private: 55 static const int kMaxSampleRate = 48000; 56 static const size_t kExpandDownsampLength = 100; 57 static const size_t kInputDownsampLength = 40; 58 static const size_t kMaxCorrelationLength = 60; 59 60 // Calls |expand_| to get more expansion data to merge with. The data is 61 // written to |expanded_signal_|. Returns the length of the expanded data, 62 // while |expand_period| will be the number of samples in one expansion period 63 // (typically one pitch period). The value of |old_length| will be the number 64 // of samples that were taken from the |sync_buffer_|. 65 size_t GetExpandedSignal(size_t* old_length, size_t* expand_period); 66 67 // Analyzes |input| and |expanded_signal| and returns muting factor (Q14) to 68 // be used on the new data. 69 int16_t SignalScaling(const int16_t* input, 70 size_t input_length, 71 const int16_t* expanded_signal) const; 72 73 // Downsamples |input| (|input_length| samples) and |expanded_signal| to 74 // 4 kHz sample rate. The downsampled signals are written to 75 // |input_downsampled_| and |expanded_downsampled_|, respectively. 76 void Downsample(const int16_t* input, 77 size_t input_length, 78 const int16_t* expanded_signal, 79 size_t expanded_length); 80 81 // Calculates cross-correlation between |input_downsampled_| and 82 // |expanded_downsampled_|, and finds the correlation maximum. The maximizing 83 // lag is returned. 84 size_t CorrelateAndPeakSearch(size_t start_position, 85 size_t input_length, 86 size_t expand_period) const; 87 88 const int fs_mult_; // fs_hz_ / 8000. 89 const size_t timestamps_per_call_; 90 Expand* expand_; 91 SyncBuffer* sync_buffer_; 92 int16_t expanded_downsampled_[kExpandDownsampLength]; 93 int16_t input_downsampled_[kInputDownsampLength]; 94 AudioMultiVector expanded_; 95 std::vector<int16_t> temp_data_; 96 97 RTC_DISALLOW_COPY_AND_ASSIGN(Merge); 98 }; 99 100 } // namespace webrtc 101 #endif // MODULES_AUDIO_CODING_NETEQ_MERGE_H_ 102